Automatic target matching

Many photogrametric processes require a large number of points to be collected from numerous digital images. It is imperative that these points be collected accurately, so that precise real-world coordinates may be assigned to points captured in the image. To this end, many techniques have been developed to locate, track and identify image targets. This thesis outlines many of these techniques and presents a target matching solution that has been developed in C++, for the subpixel target location program INDMET . The target matching solution is composed of three elements: an epipolar line program, a cross correlation program and a template least squares matching program. The epipolar line program is used to limit the search area in the right image of a given stereo pair, to the vicinity of a single line. The cross correlation program searches this line to locate possible targets and the template least squares matching program is used to determine the target centre of a black and white image target, once it has been located. It was found that these three programs, working together, had between a 20 and 70 percent chance of locating the correct target, depending on the similarity of elliptical targets in each image. Once found, the program could calculate the target centre to an accuracy of approximately 1/10th of a pixel.